Lateral AI: Simulating Diversity in Virtual Communities

Abstract

In this paper, we present Lateral AI that offers a diverse and multi-dimensional world experience. It makes use of semi-automated prompt engineering on top of GPT3.5. The coupling with named entity recognition and text summarization enables creation of a diversity of AI personas and a multiplicity of requests. The features of Lateral AI, such as creation of custom AI personas, prioritisation of user-embedded knowledge in those personas and follow-up requests, enable users to co-create with AI. Users can contribute certain information and perspectives to the application if a Large Language Model does not have access to it. Lateral AI makes the user an active component of the integrated system rather than a mere AI consumer. We demonstrate use of Lateral AI to generate a range of diverse responses and illustrate the ability of AI to predict beyond its factual knowledge. Lateral AI is a unique and alternative option to other AI models, contributing to the diverse and creative pool of emerging AI technologies and applications. The principles behind Lateral AI can be used to simulate diverse communities in a variety of settings such as online virtual communities and human robotics.